ACEEE Int. J. on Network Security , Vol. 02, No. 04, Oct 2011
A Social Welfare Approach in Increasing the Benefits
from the Internet in Developing Countries
Elly A. Gamukama1, Oliver B. Popov2
1,2
Department of Computer and System Sciences, Stockholm University
Forum 100, Isafjordsgatan 39, SE-16440, Kista, Sweden.
Email: {gamukama, popov}@dsv.su.se
Abstract— The paper examines the Internet usage and its
market environment in developing countries under the
perceived assumption that the Internet is one of the most
important drivers for development. It gives an insight on
processes’ (both unintended and intended) implications and
their effects on achieving real Internet benefits in the
environments where network infrastructures are limited such
as the ones found in the developing regions. A welfare based
approach is proposed in which the Internet providers and endusers identify a set of objective that leads them in achieving
increased benefits. Analytical model of the main
characteristics in the approach is presented and eventually
shown how the end user bit rate could be regulated based on
the utility bounds that lead general satisfaction to all users.
User satisfaction signifies delivery of expected QoS and as
well as willing to pay for such services.
There are well established findings showing that
protocols insensitive to fair sharing of network resources
among users may lead to flow starvation and system
imbalances which may further deteriorate QoS delivered. The
later phenomenon causes user dissatisfaction and creates a
negative perception of the Internet potential to development.
In regions with limited network infrastructures the usual
solution has been “more money for capacity increase”. As
one of the means for profit maximization this has arguably
denied a large number of low-income earners (and in particular
those from rural areas) from accessing the Internet and its
benefits. One of the possible ways to remedy the situation is
the societal approach in which the Internet providers and
end-users identify a set of objective that lead to optimal
benefits. The discourse reflects the present nature of the
Internet hierarchical market in developing countries. The
Internet market is viewed as a 3-level hierarchy, whereby the
topmost level is the Network Provider (NP), middle is the
Internet Service Provider (ISP) that normally sells the
broadband to the third/bottom level who is the end-user. In
this context the end-user can be a corporate or a residential
consumer. In view of the liberalization of the telecom sector
in developing countries due to the lack infrastructures, most
private companies have setup their own to provide the
Internet services. Therefore, our approach combines the roles
of NP and ISP into one termed as the Network Service Provider
(NSP). Under such a market setup, we conceptualize a societal
approach based on the concept of fairness and efficiency
that leads in establishing optimal point in achieving the
benefits from the Internet delivered services that could foster
development (a hypothetical thought that a higher
penetration of Internet can influence many aspects of life: economic productivity, health, education, democracy, etc.).
We quantify the forethought benefits of NSP from the point
of profit maximization. While the end-user (or the Internet
consumer) benefits are modelled from utility maximization of
the connectivity paid for in respect of ones preferential needs.
A tradeoff analysis among these concepts through both
analytical proofs and simulation experiments establishes a
feasible region that guarantees delivery of selected services
considered essential for fostering development. The gist for
the efficiency-fairness tradeoff presented within this
approach is to control the distortion of fairness concept in IP
based capacity-constrained networks, which is essential in
developing regions where there is deficiency in network
capacity. Therefore, the goal is to parameterize tradeoff
between services equality (sensitivity to service fairness) and
Index Terms—welfare, benefits, fairness, Internet, user
satisfaction, social-economic progress
I. INTRODUCTION
Information and communication technology and
development (ICTD) is a research area that has broadly
captured the attention of the public and the academics in the
last two decades. It deals with the interaction and the relations
between the humans and the society in general on one side
and the technology on the other side. The focus is on the
computing and communication technology, including the
Internet, which is seen as one of the enablers for economic
and social growth. The benefit of the Internet connectivity
and usage in inducing and enhancing positive social changes
in basic dimensions1of human life is generally accepted as
one of the most important drivers for development [1, 2].
Hence the success and the inevitability of the Internet in the
developed world underline its proliferation and diffusion
essential in less developed countries. However, sometimes
these processes are being impaired by unintended and
intended consequences created by the social dynamics
coupled with market environments. This paper attempts to
give an insight in both unintended and intended implications
and their effects in the environments where network
infrastructures are limited such as the ones found in the
developing regions. Examples of the former (unintended) are
the protocols being used and the applications being preferred,
while insistences of the later (intended) are the objectives of
the network providers for profit maximization.
1 Basic dimensions = economic productivity, health, education,
democracy, quality of life
© 2011 ACEEE
DOI: 01.IJNS.02.04. 55
29
ACEEE Int. J. on Network Security , Vol. 02, No. 04, Oct 2011
throughput maximization (sensitivity to effort fairness). This
address the issue of letting the users to go by there preferences
in using different applications, but control the possible
distortion such that the value “utility” end users get are fairly
equitable. Similarly, sensitivity to effort fairness addresses
the issue of throughput maximization which is the cornerstone
of NSP concerns. The rest of the paper is organised as follows;
Section II gives an insight on the unintended and intended
consequences as a result of the Internet usage vis-à-vis its
market environment from an end user social dynamics and
NSP profit maximization perspectives. Section III discusses
the societal approach based on the welfare methodologies
that would increase the benefits for both parties (end users
and NSP). Section IV, gives a conclusion of the paper and the
future outlook.
II.
resource. In most of the constrained networks, Internet traffic
is predominantly characterized by a so-called best effort
approach. In this approach the end user´s traffic is basically
handled by the end-to-end mechanism. That is, the chains of
network elements are left to handle the Internet traffic as
efficiently as possible, but without any form of guarantee.
Depending on the structure of the networks and their
elements, instantaneous number of traffic flows, their
characteristics, time of the day, etc., different end users may
experience different network performance. In case a network
is congested, then the anticipated network performance leads
to disutilities and dissatisfaction of end users. We classify
origins of such anticipated network performance in the present
Internet architecture into three main domains as consumer
(end user), NSP, and producer (content or service provider).
1) In the consumer (end user) domain
Customer premise equipment: this is a central component
in the end users environment. It has an influence on the end
user’s perception on the quality of the connection as it
connects to the NSP equipment. This is mainly due to
hardware, software, power levels of the equipment, and the
connection medium. Connection medium: it makes a difference
whether the end user is connected via an Ethernet, fiber links,
or wireless via a WiFi, WiMAX or Bluetooth. It addition it
matters whether other people are sharing the same home
network and transport their traffic simultaneously over the
same links/channel. User preferences and behaviour usage:
This is a component that greatly attribute to negative
perceptions to end users mainly in weak infrastructure
networks. An end user may intentionally or unintentionally
start an application to access ones Internet based service.
Depending on the application design and its nature of the
underlying network transport protocol, the application may
start multiple flows that results in congesting the network.
2) In the NSP domain
Offered maximum bandwidth: this is the major factor that
determines end-to-end transmission rate. Besides the more
direct technical factors related to the type of access network
used (fixed or mobile), the actual network centric measure
(packet loss, delay and response time) and the instantaneous
number of users online sharing the using the same medium
(network link or the same frequency space) are also of grate
importance. If the NSP is more driven by profit maximization
and does not offer preferential treatment to end users, then
network performance may lead to different perception among
end users for the same Internet access subscription,
depending on time and place. Differentiation between
downstream and upstream: In capacity constrained networks
that tend to offer flat pricing on shared medium, NSP tend to
offer high download stream capabilities at the expense of the
upstream capabilities. This phenomenon ends up giving end
users ford of uploading data a different perception of network
performance from those one that are merely
consumers.Network peering and Data routing: local NSP peer
with other networks. The traffic of the end user might be going
beyond the local/national network provider. Consequently
the NSP has no control on delays in another network domain.
CONSEQUENCES OF THE INTERNET SOCIAL DYNAMICS VIS-À- VIS
MARKET ENVIROMENTS
A. Unintended and Intended Consequences
In literature with respect to innovation and diffusion of
technology, the term “consequences” refers to the changes
that occur to an individual or to a social system as a result of
adoption or rejection of a technological innovation.
Consequences may take either or a combination of the
following forms; desirable or undesirable, direct or indirect,
and anticipated or unanticipated [3]. Whichever outcome may
be, the sole goal to introduce innovations to a client system
is always in good faith expecting the desirable, direct, and
anticipated consequences thought it is not always the case.
In lieu of the liberalization/privatization of the telecom market
in most of the developing countries, among others, states
anticipated the availability and affordability of the Internet
access that provides the end user (citizens) with a global
access to Internet based services. So that such a global
access would open new opportunities that accrue positive
benefits. While the NSP anticipated positive growth and
increased income from the Internet access provision
business. Generally it should be noted that unlike in
developed countries where the introduction of the Internet
was mainly a result of the urge to ease communication
between research groups, in developing countries it was
purely introduced for commercial gains. In this context the
end user was seen are a consumer. Although over a decade
passed now an end user can be a consumer or a producer of
content or/and online services. Ideally the preferences of an
end user are basically driven by several attributes in choosing
the NSP. Some of the main attributes can be enumerated as
cost, speed, and reliability. But the sole goal for paying the
price for Internet access is to maximize its utility. In this
context, the utility expresses the extent to which one can
achieve the maximum benefit expected from ones’ preferred
services that compensates for the expenditure made. However,
it is important to note that common end users only afford to
pay for share medium. The resource (i.e. the bandwidth) is
shared among the several end users. As each link capacity is
finite, the end users are essentially competing for the scarce
© 2011 ACEEE
DOI: 01.IJNS.02.04.55
30
ACEEE Int. J. on Network Security , Vol. 02, No. 04, Oct 2011
Even if the local NSP offers a higher speed, there will be packet
drops or delays at the peering router that will result in different
perception of network performance to the end user.
Consequently the only alternative the NSP could do is to
provider another route if available. But such a route might be
more expensive hence frustrating the end user too.
1) In the producer domain
The perception of the QoS offered over the Internet also
depends strongly on how the servers hosting the Internet
page or Internet services are dimensioned, and how they are
connected to the Internet. For example a transcontinental
access of a web page or Internet service is likely to deliver
low QoS if not blocked to an end user in developing countries
considering the anticipated network influences as discussed
in the above two domains. In addition some online services
are unavailable or blocked from some users because of license
for use in a certain geographic area. OR the cost of the
bandwidth for accompanying adverts turns not to be cost
effective beyond a given number of hopes in the network or
beyond a targeted region. Although some large content and
service online services providers have sent up their our
backbone network in different regions of the world, end users
still can perceive different QoS depending on ones’ last-mile
link.
A remedy to this situation, we classify the Internet
services by their traffic class. Each traffic class has a network
resource demand that is reflected by its utility function. We
categorize services as shown in Table 1 below based on their
network resource demands and respective priori thought of
importance/relevance to low-income earners. The Service
Relevance (SR) is ranked on the scale of 1 (Highest) to 5
(lowest) that indicates the deemed importance of the service
to development. Under the assumption that the user utility
of each traffic class is determined only by the allocated
equivalent bandwidth
, we introduce a new parameter
into the utility function.
T ABLE 1
TRAFFIC CLASSIFICATION AND THEIR RELEVANCE
(Details of
B. Social Dynamics and Internet Services
However taking the considerations of the market
environment in the study area and the nature of the commodity
in trading, we note that the Internet is not like other types of
production services where producers are expected to compete
for customers only through their choices of price and types
of services. The Internet market has another force hereby we
refer to as “social dynamics1” of the Internet, which is a
consequence of the rapid technology innovations and user
generation age gap. The social dynamics introduce
applications on the network that have remarkable distinction
of quality of services (QoS) requirements. NSPs face a
challenge of providing such services that “we classify as
secondly in context of development but primary in
maintaining some of the clients demands/preferences despite
the fact that they glossily contribute to congestion over the
weak infrastructure. However, failure to meet such demands
cause looses of clients to the NSP. For example we consider email or/and VOIP as one of the primary class services in context
of development. While Internet gaming or/and IPTV as in the
secondary class. But to an NSP whose billing system is based
on usage/volume-billing rather than rate-billing will gain more
profits from the secondary class clients. Such an incident
takes the efficiency-fairness tradeoff outside the feasible
region that carter for all social class of users (haves and havenots). Consequently the NSP goals of profit maximization on
weak infrastructure are misaligned with the goals of Internet
provision for development.
see annex A)
III. A SOCIETAL APPROACH FOR INTERNET ACCESS
A. Social goal of equal utility
The goal of equal utility is that all users should get the
same level of satisfaction from the network regardless of the
different services in their respective traffic classes. The
framework adopts the concept of utility fair as criteria for
capacity allocation in networks [6]. A point to observe is that
allotting equal utilities to all end-users across different types
of services in their respective traffic classes does not mean
that they were allocated equal bandwidth across their different
traffic classes. Users within the same traffic class have
identical utility functions, thus the utility-fair bandwidth
2 In this context “social dynamics” refer to the currently shift or move from the use
of the traditional mode of services delivery to the Internet or IP based service delivery
mode. On a close view, we observe that there is an explosion in voice and video
applications that are constantly being used to deliver services to the end users. i.e.
from the traditional telephony systems to VOIP base (e.g. Skype, messenger, etc.).
TVs and Movies are constantly being watches on the Internet.
© 2011 ACEEE
DOI: 01.IJNS.02.04. 55
gives the weight of
a traffic class flow based on its deemed level of importance
vis-à-vis other traffic classes in the network.
Further we characterize the networks from the point of striving
to deliver services that meets user satisfaction levels while
maintaining the tradeoffs within the feasible region. User
satisfaction is a standard measure of network performance
[4] and user willing to pay for the service received from
microeconomics point of view.
31
ACEEE Int. J. on Network Security , Vol. 02, No. 04, Oct 2011
allocation inside each class of user group also leads to the
bandwidth-fair allocation that is within [
,
] rage
for each class. Refer to annex A for details. Every traffic class
has a required minimum bandwidth denoted as
, and a
maximum threshold bandwidth denoted by
beyond
which any allocation does not benefit the user/application.
Based on clear and fair decisions, the NSP normalizes the
traffic taking considerations of the importance of various
traffic classes and user benefits. It is done by getting a product
of the scale
and traffic class utility function. Then actual
utility of the end user can be quantified. We elaborate this
characteristic through an example below. Demonstrating the
concept of social goal of equal utility across users irrespective
of the class type of service is our starting point. Consider a
network consisting of a single link with a unity capacity being
shared by two users s1 and s2. User s1 transfers data according
to an elastic application with strictly increasing and concave
bandwidth utility µ1(x1). User s2 transfers real-time video
data with a non-concave bandwidth utility µ2(x2). As shown
in Figure 1 below, if the link capacity is shared equally using
commonly know max-min bandwidth allocation policy, each
user gets equal bandwidth of x = 0.5units, with corresponding
utility of µ1(x = 0.5) and µ2(x = 0.5) respectively. It is obvious
that for user s2 is not satisfied because the received bandwidth
does not reach the minimum required level for video encoding.
NSP and traffic class function. Through this normalization
then the actual utility of the end user can be quantified.
B. Utility based network modeling
We follow the works of Low and Lapsley [9], in which
social utility was maximized by a bandwidth pricing algorithm.
For a single link the optimization problem is as follows:
In such a case the maximum total utility is thought as being
the sum of all utilities
that each user associates with
ones’ allocated bandwidth
under link capacity
constraint . The users’ utility is strictly concave and increase
with the allocated bandwidth that is bounded in the rage.
Beyond the utility marginal increase
is almost zero and below the users’ utility becomes zero. The
centrality of optimization approach above and its solutions
delivered through the Lagrange dual problem decomposition
methods enable the network system (in this context the NSP)
to establish the aggregate link price
at optimal bandwidth
allocation levels. Then price is passed over to the user who
decides on what to purchase in view of what maximizes ones’
objectives. The resulting net-utility of the user is normally
formulated in the form of
Really this problem
solving is to let the user choose what one can afford to pay
for in lieu of ones’ financial possibilities. Unlike the utility
optimization form above, we propose a model that guides the
user on what to take based on user preferences in lieu of
services importance in context of development1. We assume
that each service has a particular importance to its user
relatively to ones’ preference. In addition the cost of network
resource requirements for such a service should be in the
rage of users’ capacity to pay. On this basis we introduce the
user preference parameter in the optimization that describes
the service utility (importance degree) in fostering
development. Flows are differentiated based on the deemed
importance degree of the services they deliver in accordance
to fostering development. We introduce a scalar value
in
the utility function that depicts the traffic class importance
and the required QoS levels according to the user service
preferences.
Figure 1: Utility max-min and bandwidth max-min fairness
Likewise, if the utility allocation policy ensures that the two
users get equally utility, i.e. µ1(x1) = µ2(x2) = µ*, then
accordingly both users are satisfied, hence willing to pay for
the services. The latter allocation policy represents the utility
max-min fair. The utility max-min algorithm was introduced in
[7] by Cao and Zegura and its architecture in [8] by Cho and
Chong. The social goal of equal utility across users is
irrespective of the class type of service. A point to continue
being observed is that equal utilities to the end-users across
different types of services do not mean equal bandwidth to
different classes or even to users within the same class type
of service. As indicated in Table 1 above, all the traffic classes
have a required minimum capacity/bandwidth other than the
“Elastic” that can start from zero upwards. Similarly every
traffic class has a maximum threshold bandwidth, beyond
which any allocation does not benefit the user/application.
Therefore the NSP, based on clear and fair decisions, has to
normalize the traffic taking considerations of the importance
of various traffic classes and user benefits. The normalization
is done by getting a product of the scale determined by the
© 2011 ACEEE
DOI: 01.IJNS.02.04.55
IV. CONCLUSIONS AND FUTURE WORK
In this paper we have presented a social welfare approach
that characterizes what would be the environment of the
Internet services in view of in view to raise the benefit of the
end users and NSP developing regions. Analytical model of
the main characteristics in the approach is presented and
eventually shown how the end user bit rate could be regulated
based on the utility bounds that lead general satisfaction to
all users. At individual level we shall advance our study based
on the established utility levels that deem to deliver user
satisfaction for a given traffic class. User satisfaction signifies
32
ACEEE Int. J. on Network Security , Vol. 02, No. 04, Oct 2011
delivery of expected QoS and as well as willing to pay for
such services. Hence the focus remains on testing the actual
benefits that foster a positive social change in one’s basic
dimensions of life from those traffic classes with high SR. We
adopting the benefit function in [10], where individual benefits
are computed and their aggregation lead us in achieving the
group benefit. Here-forth referred to as the society’s welfare.
[2] M. Warschauer and T. Matuchniak, “New Technology and
Digital Worlds: Analyzing Evidence of Equity in Access, Use, and
Outcomes,” Review of Research in Education, vol. 34, pp. 179225, March 2010.
[3] E. M. Rogers, Diffusion of Innovation, 5th ed.: Free Press,
2003.
[4] S. Shenker, “Fundamental Design Issues for the Future
Internet,” IEEE Journal on selected area in Communications, vol.
13, pp. 1176-1188, 1995.
[5] L. Shi, C. Liu, and B. Liu, “Network Utility Maximization for
Triple-play Services,” Computer communications, vol. 31, pp. 2257
- 2269, 2008.
[6] T. Harks, “Utility Proportional Fair Bandwidth Allocation An optimization Oriented Approach,” August 2004 2004.
[7] Z. Cao and E. W. Zegura, “Utility Max-Min: An ApplicationOriented Bandwidth Allocation Scheme,” 1999, pp. 793-801.
[8] J. Cho and S. Chong, “Utility Max-Min Flow Control Using
Slope-Restricted Utility Functions,” Communications, IEEE
Transactions on, vol. 55, pp. 963-972, 2007.
[9] S. H. Low and D. E. Lapsley, “Optimization flow control I:
Basic Algorithm and Convergence,” IEEE/ACM Transactions on
networking, vol. 7, pp. 861-874, 1999.
[10] D. G. Luenberger, “Benefit functions and duality,” Journal of
Mathematical Economics, vol. 21, pp. 461-481, 1992.
ACKNOWLEDGMENT
This work is supported by the Swedish International
Development Agency through her Department for Research
Cooperation (Sida/SAREC) under the framework for
strengthening national research capacity in developing
countries. We gratefully acknowledge Sida/SAREC.
REFERENCES
[1] R. Gregory, S. Scott, and W. Donald, “Household demand for
broadband internet service,” Communications of the ACM, vol. 54,
pp. 29-31, Febuary 2011.
ANNEX A ( THE
PROOFS OF THE EQUATIONS REFER TO
3 Implication: the higher the importance of a service is deemed to
foster development, the lower the price for the capacity/bandwidth
to deliver it to the user.
© 2011 ACEEE
DOI: 01.IJNS.02.04. 55
33
[4])